Low Power, Wide Area Networks

Low power wide area networks (LPWAN) are not a new phenomenon HOWEVER, THEY ARE BECOMING MORE POPULAR DUE TO THE GROWTH OF THE INTERNET OF THINGS (IoT).

LPWAN is often used when other wireless networks aren’t a good fit—Bluetooth and BLE (and, to a lesser extent, Wi-Fi and ZigBee) are often not suited for long-range performance, and cellular M2M networks are costly, consume a lot of power, and are expensive as far as hardware and services are concerned.

LPWAN technology is perfectly suited for connecting devices that need to send small amounts of data over a long range, while maintaining long battery life. Some IoT applications only need to transmit tiny amounts of information—a parking garage sensor, for example, which only transmits when a spot is open or when it is taken. The low power consumption of such a device allows that task to be carried out with minimal cost and battery draw.

LPWAN FEATURES

1.      LONG RANGE: The end-nodes can be up to 10 kilometers from the gateway, depending on the technology deployed.

2.      LOW DATA RATE: Less than 5,000 bits per second. Often only 20-256 bytes per message are sent several times a day.

3.      LOW POWER CONSUMPTION: This makes very long battery life, often between five and 10 years, possible.

There are two main areas where LPWAN technologies are best suited

1.      FIXED, MEDIUM- TO HIGH-DENSITY CONNECTIONS: In cities or buildings, LPWAN technologies are a great alternative to cellular M2M connections. Some examples include smart lighting controllers, distribution automation (smart grid), and campus or city focused GPS asset tracking.

2.      LONG LIFE, BATTERY POWERED APPLICATIONS: When a longer range is needed than legacy technologies can provide, LPWAN can be a good fit. Examples include wide-area water metering, gas detectors, smart agriculture, and battery powered door locks and access control points.

THE SWEET SPOT FOR LPWAN

Different wireless technologies address application-specific needs with changes in modulation and frequency schemes. Long-range applications with low bandwidth requirements that are typical for IoT applications are not supported well by these existing technologies.

Network Configuration

MESH TOPOLOGY: Many LPWAN customers have previously tried to solve their wireless connection problems with mesh topology networks like ZigBee. They struggled with mesh network solutions because the link budgets for these connections are very limited due to high data rates and low receiver sensitivities. Some ZigBee connections have trouble sending data more than 20-30 meters away because the power coming from the transmitter is lost too quickly. Additionally, others have been surprised by the amount of mesh infrastructure required to build a reliable network.

STAR TOPOLOGY: Instead of a mesh topology network, most LPWAN technologies use a star topology network. Like Wi-Fi, the endpoints of star networks are connected directly to the access point. Link Labs can use a repeater to easily fill in gaps in coverage, which, for most applications, is a good middle ground in terms of latency, reliability, and coverage.

Fundamental LPWAN Concepts

Range Vs. Data Rate: TO ACHIEVE LONG RANGE IN WIRELESS COMMUNICATIONS, YOU NEED A LARGE LINK BUDGET.

In other words, when you transmit a signal, it needs enough energy to be detected when it’s received. Because a certain amount of power is lost along the way as it propagates through space and materials in between, there is a baseline amount needed to transmit the signal properly.

LPWAN technologies generally operate with about 140-160 decibels (dB) of total path, which can add up too many miles of range in the right circumstances. This is primarily achieved by high receiver sensitivities. Receiver sensitivities of more than -130 dBm are common in LPWAN technologies, compared with the -90 to -110 dBm seen in many traditional wireless technologies. Technologies with -130 dBm can detect signals 10,000 times weaker than technologies with -90 dBm, so you can see how this is important for LPWAN.

The slower the modulation rate, the higher the receiver sensitivity can be. This comes down to the Shannon-Hartley theorem, or Information Theory, which states that the energy per symbol or energy per bit is the main lever to change the possibility of a message being heard. By slowing the modulation rate by half, you are putting twice as much energy into each symbol; thus, you are increasing the link budget, or receiver sensitivity, by double (3 dB).

Sigfox1 is an example of how modulation rate and range are connected. Sigfox transmits data using a standard radio transmission method called binary phase-shift keying (BPSK). Its modulation rate—300 bps—is extremely slow in a modern sense. But due to this slow modulation rate, it’s able to get great range with fewer base stations.

In the U.S., Sigfox modulates at a higher rate, because otherwise it would not be able to meet the FCC Part 15 requirement that the maximum time a transmission can be on the air is 0.4 seconds.

Processing Gain

THE TECHNICAL DEFINITION FOR PROCESSING GAIN IS THE RATIO OF THE RADIO FREQUENCY (RF) BANDWIDTH TO THE UNSPREAD BANDWIDTH, USUALLY EXPRESSED IN DECIBELS.

Here’s a simple way to think of it: imagine you’re sitting in front of a TV screen, and all you see is static. That static can be thought of as a visual representation of noise. Now let’s assume you can press pause on your TV remote, freeze the static, put a transparency to your TV screen, and color in all the black pixels until you had an exact replica of the static at that moment. If you then decided to label this transparency as “Static X,” you could press play again and, with the transparency in hand, watch the static until you saw a frame that looked like your drawing. Once this happened, you could say that someone had transmitted Static X.

When applied in more realistic terms, this processing gain illustration shows that when a signal is mixed across the RF spectrum, it is only detectable when you have processed all the noise and are looking at it with a filter. Negative dB signal-to-noise ratio (SNR) means that the signal is below the noise floor: it can’t be seen with a simple receiver unless you are looking for it. This, in a nutshell, is processing gain.

As another example, Sigfox technology is a BPSK narrowband signal with very narrow channel sizes. Weak signals are more easily detected in a narrow channel, since the noise floor is effectively lower than it is for wider band signals like LoRa. This is because noise is spread throughout the spectrum. If your receiver bandwidth is smaller, then the noise level is smaller too. However, traditional frequency-shift keying (FSK) signals— which transmit information through the frequency changes of a carrier wave—have no “processing” or “coding” gain. This means they must have a positive signal-to-noise ratio of around 10 dB to detect the signal.

When coding is used, a signal can be detected up to around -20 dB SNR. For code-division multiple access (CDMA) signals like Ingenu’s or chirp spread spectrum (CSS) modulations like LoRa, the effects of a higher receiver noise floor are mitigated by processing gain. For the most part, coded signals are better than narrowband signals in terms of minimum detectable energy, but there are some drawbacks associated with them, which we will discuss below.

Noise Vs. Bandwidth

AS REFERENCED DURING OUR DISCUSSION ON PROCESSING GAIN, THE NOISE FLOOR OF A RECEIVER IS SET BY TWO THINGS: THE BANDWIDTH AND THE NOISE.

Think of it this way: if you look through a pinhole, you see less light than if you look through a paper towel roll. That same logic can be applied to radios.

A narrowband channel of 100 Hz has a thermal noise floor of about -154 dBm, which means that if you require a 10 dB SNR, your ideal receiver (and theoretical maximum sensitivity) would get -144 dBm (unless you use coding). If you use coding in that channel, then you can get below the noise floor to the same sensitivity—but coding a signal requires bandwidth to spread the energy across.

Interference 

BASED ON THE COMPARISON BETWEEN NOISE AND BANDWIDTH, YOU UNDERSTAND THAT THEORETICAL PERFORMANCE BETWEEN A NARROWBAND CHANNEL AND A CODED CHANNEL IS THE SAME.

But many people in the LPWAN space disagree on which technology is better when it comes to noise. (For example, this article demonstrates Texas Instruments’ opinion3 on long-range RF communication.)

Narrowband noise can be thought of as actual narrow signals sitting above the noise floor. If you’re a narrowband system (like Sigfox) looking at a tiny 100 Hz channel, the channel next door can be loud, and it still won’t affect you. If it’s an interferer in the same channel, however, you’re going to get clobbered.

For narrowband noise, which much of the 900 MHz ISM band interference is, if a narrowband signal gets “clobbered” (i.e., a signal lands right in the channel), it has very poor blocking performance. However, narrowband systems can have more than 50 dB adjacent channel rejection.

Wide-band noise is like widespread noise that effectively raises the noise floor. Narrowband interference is less of a problem for a wide-band, coded system, since it just adds to the overall noise in the band.

Wondering which system is better? Frankly, we don’t know, because it depends on the specifics of the environment. This is widely discussed within the LPWAN space, but we won’t make any speculations within this paper.

Licensed Vs. Unlicensed

MOST CURRENT LPWAN TECHNOLOGIES USE AN UNLICENSED BAND. SIGFOX AND LINK LABS BOTH USE THE 900 MHZ ISM BAND IN THE U.S. AND THE 868 MHZ BAND IN EUROPE. INGENU USES THE 2.4 GHZ BAND.

All the technologies mentioned above work just as well in licensed bands. In fact, they work better because there is less interference from other users. So, what’s the issue? When using licensed bands, you have to re-tool the MAC scheme to deal with different channel size, spacing, etc. For instance, if you go to a licensed spectrum, you’d probably have less than 1 MHz of spectrum, whereas in unlicensed bands, you could get 26 MHz.

Both coded and narrowband signals work well in licensed spectrums but using them becomes an issue of spectral efficiency and capacity more than anything else. (We’ll discuss this in depth in the section on orthogonality). Essentially, the challenge is to pack as much data flow into the band as possible. The FCC Part 15 and ETSI rules go out the window as well, because as a license holder, you have much more freedom to use your spectrum to your advantage.

However, the issue of licensed vs. unlicensed spectrums may not be an issue much longer. In August 2015, the GSMA (Groupe Speciale Mobile Association)—a group made up of mobile operators—announced that it plans on standardizing LPWAN technology on a licensed spectrum by early 2016. This push has received endorsements and backing from companies like “AT&T, Bell Canada, China Mobile, China Telecom, China Unicom, Deutsche Telekom, Etisalat, KDDI, NTT DOCOMO, Ooredoo, Orange, Singtel, Telecom Italia, Telefonica, Telenor, Telstra and Vodafone,” according to an article put out by Telecom TV. We’re very interested to see if this breaking LPWAN news moves forward and on GSMA’s proposed timeline.

Sub-GHz Spectrum Availability Worldwide

THIS IS PROBABLY THE BIGGEST DRAWBACK FOR LPWAN TECHNOLOGIES IN THE 900/868 MHZ BAND.

Every country has different rules about using the sub-GHz spectrum. There are generally two camps: those that follow Europe (868 MHz), and those that follow the U.S. (915 MHz). The 915 MHz band is available only in about a third of the world, and some countries don’t have any bands available. In fact, many countries have added special caveats that make standardization nearly impossible. Until this issue is resolved, there is no globally available band for LPWAN technologies like there is at the 2.4 GHz level (for Bluetooth and WiFi).

Localization Capabilities

MEASURING THE LOCATION OF AN RF SIGNAL IS DONE BY ESSENTIALLY CONVERTING THE TIME OF ARRIVAL INTO A DIRECT PATH LENGTH. SO, TO MEASURE TIME, YOU ABSOLUTELY MUST BE ABLE TO DETECT A DIRECT PATH.

There are two things that go into the location (or really the time of arrival) of an RF signal: enough power to detect the direct path and enough bandwidth to resolve the multipath reflections from the direct path.

Imagine you’re in the living room, and someone is in the bedroom with a strobe light on. You can’t see this person, but you can see the strobe lights, because the light is bouncing around and refracting off the walls. In a discussion on network localization capabilities, that type of light transmission would be considered a multipath (or non-direct) path.

Most LPWAN technologies are not received on the direct path; they are received on the multipath channels. This is a good thing for data reception, since weak signals that have bounced frequently are still received, but it also means that LPWAN technologies are not ideal for localization. Unfortunately, no amount of averaging can change this—it’s simply the laws of physics. Averaging only helps if something is moving (in space and frequency), and in most LPWAN systems, neither of these are happening.

On the other hand, if you’re using radio waves, you must be able to detect the direct path. If you’re unable to do so, you’ll be potentially creating a huge area of uncertainty. If you used the wrong measurements in your calculations, for example, you could end up with a kilometer-wide uncertainty circle.

Signal bandwidth is required because the ability to determine the difference in path length between two signals (say, the direct path, and a multipath reflection) is a function of signal bandwidth. A narrowband signal (100 Hz) could never be used for accurate time-based measurement, and even a 125 kHz LoRa signal only has a multi-path resolution ability of about 1 km. That means if there are any reflected paths with a length of less than one kilometer different from the direct path, the measurement will not be accurate. Because of these limitations, we encourage those who need localization capabilities to investigate GPS, WiFi, or proximity-based RFID.

Orthogonality

WHEN TWO LINES ARE ORTHOGONAL, IT MEANS THEY ARE BOTH RIGHT ANGLES. THE RF WORLD “HIJACKED” THIS TERM, IF YOU WILL, TO MEAN TWO SIMULTANEOUS SIGNALS THAT ARE BOTH DETECTABLE.

So, orthogonality is detecting multiple data streams in the same channel and at the same time. This is a feature of a coded channel, and it offers a solution for getting back good spectral efficiency for wider band systems. Because coded signals are spread across a larger swath of spectrum, those signals take up more frequency real estate. Narrowband signals, however, can pack quite a bit of traffic into that same bandwidth. If there are multiple coded streams simultaneously on the air, you buy back some (though usually not all) of the spectral efficiency you give up with coding. FSK systems cannot detect more than one signal at a time, and if two signals use the same channel at the same time, only the stronger signal will be decoded, if at all.

Importance of MAC Protocols

Ultimately, much of the value of a LPWAN technology is not the underlying RF characteristics, if the “link is closed.” The ability to create a network, control it, and offer bi-directional data flow is what matters most for end users. The limitations or features of one MAC implementation compared with another are very important to understand.

HERE ARE SOME LPWAN FEATURES THAT SOME LINK MAC LAYER PROVIDES:

o  Uses repeaters to fill in coverage

o  Real Time Adaptive Data Rate

o  Open Standard

o  International Roaming Support (Multi-Band)

o  100% Acknowledged Messages

o  Over-the-air Firmware Upgrades

o  Multicast Message Groups

o  Flexible Downlink Capability

o  Scalable Capacity

o  Low Downlink Latency

o  Uplink Power Control

o  Real Time Quality of Service

o  Handover

o  Interference Avoidance

o  Supports Internet Disconnected Operations

o  Supports high jitter (SATCOM) connections

o  Supports 1W Uplink Transmissions under FCC

o  Uplink-Downlink Collisions Prevented

o  MAC Layer Packetization and Retry

o  Fixed MTU Size

In Conclusion

Low power wide area networks will continue to revolutionize the way companies do business by allowing them to collect data and control devices in ways that were economically impossible before. As the technology companies described in this report begin to help companies solve their customers’ problems, we’re certain more resources will be invested in the space, which will lead to further advancements in LPWAN technology and applications.